BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Designs for mixed models with binary response - Waite\, T (Southam
 pton)
DTSTART:20110809T150000Z
DTEND:20110809T154500Z
UID:TALK32287@talks.cam.ac.uk
CONTACT:Mustapha Amrani
DESCRIPTION:For an experiment measuring a binary response\, a generalized 
 linear model such as the logistic or probit is typically used to model the
  data. However these models assume that the responses are independent. In 
 blocked experiments\, where responses in the same block are potentially co
 rrelated\, it may be appropriate to include random effects in the predicto
 r\, thus producing a generalized linear mixed model (GLMM). Obtaining desi
 gns for such models is complicated by the fact that the information matrix
 \, on which most optimality criteria are based\, is computationally expens
 ive to evaluate (indeed if one computes naively\, the search for an optima
 l design is likely to take several months). \nWhen analyzing GLMMs\, it is
  common to use analytical approximations such as marginal quasi-likelihood
  (MQL) and penalized quasi-likelihood (PQL) in place of full maximum likel
 ihood. In this talk\, we consider the use of such computationally cheap ap
 proximations as surrogates for the true information matrix when producing 
 designs. This reduces the computational burden substantially\, and enables
  us to obtain designs within a much shorter time frame. However\, other is
 sues also need to be considered such as the accuracy of the approximations
  and the dependence of the optimal design on the unknown values of the par
 ameters. In particular\, we evaluate the effectiveness of designs found us
 ing these analytic approximations through comparison to designs that are f
 ound using a more computationally expensive numerical approximation to the
  likelihood.\n
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR